Research On Grid Resources Scheduling Based On Customer Satisfaction Index And Genetic Algorithm

Grid is an important information technology springing up in the world recent years. Its aim is to achieve a broad geographical distribution of the large amount of heterogeneous resources sharing. Resources scheduling is a core element of the grid research, how to reasonably allocate tasks to different resources, is an important and practical question. Many traditional research of grid resources scheduling was performance-oriented scheduling, they paid more attention to the performance (such as the optimal makespan, resources load balancing) and economic of scheduling, but few considered the quality of service scheduling. Good quality of service however is the important power to attract more users to join and use the grid, and to promote the development of the grid. Therefore, how to improve the quality of service of scheduling based on the existing scheduling algorithms is the weak side of grid resources scheduling research at the present time.After an in-depth research and analysis for the current resources scheduling algorithms, we selected genetic algorithm as the basic algorithm to improve. Genetic algorithm is a class of random search algorithm which uses the natural selection from biological and genetic mechanisms of natural for reference. Its main features are group searching strategy and information exchanging between the individuals in the group. It is appropriate to deal with the complex and non-linear issues which traditional search method hard to resolve.In this paper, we used the economic indicator of customer satisfaction index in management economics for reference, attempted to scale the quality of service of grid resources scheduling by customer satisfaction index, proposed the definition of customer satisfaction index of grid resources scheduling, then designed the structure model of the grid resources scheduling customer satisfaction index by reference to the structure model of American customer satisfaction index, at the same time gave the evaluation methods and calculation formulas. On this basis, we proposed a customer satisfaction index based genetic algorithm (CSI-GA). Algorithm uses customer satisfaction index as a basis for the operation of genetic variation, after the evolution of several generations, then get the best individual, that is, the optimal scheduling scheme. It is an implementation for the improvement of the quality of service of grid resources scheduling. At last, we used the GridSim toolkit to make a number of simulation experiments to compare CSI-GA with Min-Min algorithm and SGA. The simulation results showed that, CSI-GA algorithm can guarantee a considerable performance and economic of the scheduling, at the same time greatly enhanced the quality of service.